1. According to Tetrachoric correlation formula, the higher the P(1,1) and P(0,0), the higher the tetrachoric correlation. Given that it is much easier to classify an examinee to have mastered/not mastered, that is, the P(1,1) and P(0,0) would be high in the test-retest context, the resulted reliability would definitely high.
2. I have a question: if an item measures more than one attribute, how can we calculate the reliability using the proposed method?